[jimmaswell]

Python only lets you have one line in a lambda - the return line. In languages with nicer anonymous function support I can make an anonymous function span multiple lines and branches eg. in C# you could assign to a delegate something like myObj.stringToHash = (s) => {int x = 0; foreach(var c in s) x += ... ; return x;}

I find python's restriction in this case fairly strange. If I wanted to do that in python I'd have to just declare a function body for what I want to put in the anonymous function elsewhere, which defeats the purpose.

[sophacles]

Functions can be defined in a nested way setting up closures. It isn't that different from what you wrote to do:

    def some_method(...):
        avar = aval
        .... (some code)
        def throwaway(x, y):
            ....
            something_else(avar, x+y)
        some_obj.somefunc = throwaway

The major difference is you do the function definition just prior to the assignment. The "namespace pollution" is literally limited to the scope of "some_method", so you have an extra local var in a function somewhere. The semantics of python mean that the function is created on the stack, and a closure is created each time the some_method is called, and the closure is later GCed as any other object would be, so there is nothing fundamentally different about the "named" function, and the "anonymous" function in you example, except a single local variable.

[jimmaswell]

That's what I meant. Defining the throwaway function directly before the assignment is a good compromise at least.

[Retra]

Well, if you need multiple lines, you probably don't need your function to be anonymous. (Your example is just `sum`...)

[jimmaswell]

I've needed it many times in the past where the function I'm assigning can't be done as a one-liner like that.